a network for supporting resting-state fMRI related studies.
Correlation with Ratings
Submitted by
cmorawe
on
Hi
I would like to correlate the connectivity parameters with covariates eg Ratings. I extracted ROI signals by defining different ROIs and get zFCMaps. Which parameters/values of the output do I have to use for correlations?
If you have got the zFCMaps, you can correlate the zFC values with the ratings. One way is use DPABI->Statistical Analysis->Correlation Analysis to do it.
In such a case, you can first organize the zFCMaps for all the subjects in the same directory. And then use DPABI->Utilities->ROI Signal Extractor to extract the zFC values of a given ROI (or several). The extracted txt file can be ported into SPSS.
I just extrated the ROI signals and got 96 values although I only have 48 subjects. What does that mean? I suspect that I have sort the data somehow.
Another concern I have: When I define the ROI with peak coordinates, eg -50 -41 12, the reported center of ROI is completley different: sphere ROI center: 46 28 27. Why is that?
When I use a mask for ROI extraction, I get the following error: Mask does not match. I got the mask using Anatomy toolbox. It should be in MNI.
One more question: When I define the output ROIs to extract the ROI signals, is it better to use masks or spheres? If I understood correctly, the output ROIs are those regions which resting state connectivity is significantly correlated with the seed regions, right? So I could use the contrast maps of the seed regions to generate masks and extract the ROI signal from the masks and correlate it with a covariate? Or is it better to determine the peak voxel and use a sphere to extract the ROI signal? Which one is the better and more commonly used approach?
I have a question: Is it better to use the FC maps or the zFC maps to do the analysis? One of my colleagues suggested that the zFC map does not reflect the actual FC values but the normalized FC values (after demean and being divided by standard deviation within each participant), and errors may be reported if we use the zFC maps to do analysis. I am looking forward to your suggestions.
If you have got the zFCMaps, you can correlate the zFC values with the ratings. One way is use DPABI->Statistical Analysis->Correlation Analysis to do it.
zFC is considered better. Here zFC means Fisher's r-to-z, not the z-standardization (subtract mean and divide by SD). Fisher's r-to-z could improve normality, thus it's recommended.
zFC is considered better. Here zFC means Fisher's r-to-z, not the z-standardization (subtract mean and divide by SD). Fisher's r-to-z could improve normality, thus it's recommended.
When I define the output ROIs to extract the ROI signals, is it better to use masks or spheres? If I understood correctly, the output ROIs are those regions which resting state connectivity is significantly correlated with the seed regions, right? This means I could either 1) generate output ROIs by using a contrast of an fMRI experiment or 2) use masks from an anatomical atlas or 3) use a sphere (coordinates from a task-related activity) and correlate it with a covariate? Which one is the better and more commonly used approach?
What is the best way to determine the sphere radius for a seed? Does it have anything to do with voxel dimensions? Currently, I am using r = 5mm because that's the value I've seen most often in the forums and literature, but the reason why 5mm has been used by other researchers is unclear to me. Any recommendations for articles or chapters that explain how to choose a seed radius would be greatly appreciated.
Hi Carmen,
Hi Carmen,
If you have got the zFCMaps, you can correlate the zFC values with the ratings. One way is use DPABI->Statistical Analysis->Correlation Analysis to do it.
Best,
Chao-Gan
Hi
Hi
Thanks for the fast reply. Where do I find the zFC values? I want to use SPSS for correlations, so I would need to export the values.
Best
Carmen
Hi Carmen,
Hi Carmen,
In such a case, you can first organize the zFCMaps for all the subjects in the same directory. And then use DPABI->Utilities->ROI Signal Extractor to extract the zFC values of a given ROI (or several). The extracted txt file can be ported into SPSS.
Best,
Chao-Gan
Hi,
Hi,
I just extrated the ROI signals and got 96 values although I only have 48 subjects. What does that mean? I suspect that I have sort the data somehow.
Another concern I have: When I define the ROI with peak coordinates, eg -50 -41 12, the reported center of ROI is completley different: sphere ROI center: 46 28 27. Why is that?
When I use a mask for ROI extraction, I get the following error: Mask does not match. I got the mask using Anatomy toolbox. It should be in MNI.
Thanks a lot. Your help is very much appreciated.
Carmen
Hi,
Hi,
1. I think you have put both zFCMaps and FCMaps to the same folder, thus there were 48*2 maps.
2. 46 28 27 is the voxel IJK index.
3. You need to resample the mask first: DPABI->Utilities->Image Reslicer.
Best,
Chao-Gan
Thanks again! :)
Thanks again! :)
One more question: When I define the output ROIs to extract the ROI signals, is it better to use masks or spheres? If I understood correctly, the output ROIs are those regions which resting state connectivity is significantly correlated with the seed regions, right? So I could use the contrast maps of the seed regions to generate masks and extract the ROI signal from the masks and correlate it with a covariate? Or is it better to determine the peak voxel and use a sphere to extract the ROI signal? Which one is the better and more commonly used approach?
Best, Carmen
Re: [RFMRI] Correlation with
Hi Delin,
Chao-Gan
Re: [RFMRI] Correlation with
ROI signal extraction and correlation with covariate
When I define the output ROIs to extract the ROI signals, is it better to use masks or spheres? If I understood correctly, the output ROIs are those regions which resting state connectivity is significantly correlated with the seed regions, right? This means I could either 1) generate output ROIs by using a contrast of an fMRI experiment or 2) use masks from an anatomical atlas or 3) use a sphere (coordinates from a task-related activity) and correlate it with a covariate? Which one is the better and more commonly used approach?
Best, Carmen
Hi,
Hi,
All the 3 methods are widely used.
1. If you have your own task data, you can define the ROIs by using the significant clusters.
2. If you have to generate ROIs form previous literature, you can generate sphere ROIs according to their coordinates.
3. If you are interested in a specific structure, you can define according to anatomical atlas.
Best,
Chao-Gan
how to determine optimal seed radius?
Hi all,
What is the best way to determine the sphere radius for a seed? Does it have anything to do with voxel dimensions? Currently, I am using r = 5mm because that's the value I've seen most often in the forums and literature, but the reason why 5mm has been used by other researchers is unclear to me. Any recommendations for articles or chapters that explain how to choose a seed radius would be greatly appreciated.
Thanks in advance,
Felix
I didn't see a clear
I didn't see a clear comparison. But 5 and 3 are mostly used.
conventional standard